# Deepseek
Company and family of Open source [[Large Language Models (LLMs)]].
Models in the family:
- R1: strongest (671B)
- R1-Zero - An R1 prototype fine-tuned using only unsupervised reinforcement learning (RL)
- VL: Vision-Language understanding
- Coder
- Coder V2
- Math
- VL
- ...
There are variants of R1, but those are distilled & fine-tuned models based on [[Qwant]], [[Llama]], etc. For example: DeepSeek-R1-Distill-Qwen-7B. That's really not R1. The only R1 is the 671B one.
Deepseek also has an API platform.
## References
- Official website: https://www.deepseek.com/
- Chat: https://chat.deepseek.com/sign_in
- Mobile app: https://cdn.deepseek.com/download-app/index.html
- Coder: https://coder.deepseek.com/
- API documentation: https://api-docs.deepseek.com/
- API platform: https://platform.deepseek.com/
- LLM source code: https://github.com/deepseek-ai/DeepSeek-LLM
- VL source code: https://github.com/deepseek-ai/DeepSeek-VL
- Coder source code: https://github.com/deepseek-ai/DeepSeek-Coder
- Coder v2 source code: https://github.com/deepseek-ai/DeepSeek-Coder-V2
- Math source code: https://github.com/deepseek-ai/DeepSeek-Math
- Deepseek @ AWS: https://aws.amazon.com/blogs/aws/deepseek-r1-models-now-available-on-aws
- Deepseek @ Azure AI Foundry: https://azure.microsoft.com/en-us/blog/deepseek-r1-is-now-available-on-azure-ai-foundry-and-github
- Videos
- Running R1 (671B) on a $2000 local server: https://www.youtube.com/watch?v=Tq_cmN4j2yY